The end of web apps

I've been making websites for more than 30 years. I've been building web apps since Single Page Applications were still new. But I'm not sure I will ever work on a web app again. Making native apps, using Claude, is now too easy.

Over the last week, I've wished 4 native iOS apps into being. One, a paid app, was published to the app store this morning. Two other free apps are mere steps away from being published. Three more are sitting on my phone, being used by me, with an eye towards publishing.

It used to take months for entire teams of highly paid and specialized software engineers to build a quality native app that was worth publishing on the app store. Authentication and authorization, highly available databases, application design, project management; all of these different functions required highly paid specialists. Coordination amongst them required even more layers, from management, human resources, recruiting, finance, security, leadership, directors, etc... It took enormous resources and planning to get even the simplest apps into the app store. Now, all an individual has to do is wish it into being.

The software genie

I'm using a combination of Hermes Agent with Anthropic's Opus, along with a couple other supporting models. Building apps from my phone is now a thing. Using Hermes on my laptop, connecting through Tailscale, means I have the full power of my desktop available to me through simple directions. I tell my agent to use my computer, or use my browser, and it can achieve anything I can.

Project Management is dead

It feels strange to think about how much time, effort, and money it used to take to actually make anything with any kind of production quality. Enormous amounts of effort were taken to scope things down to an MVP so that the large organization building the app did not waste even larger amounts of money to build something that might fail. Entire schools of thought have now become moot.

The days of scrum + agile software development is over. What's the point of trying to break down work into tickets that can be "pointed"? What is the point of arguing or even discussing "velocity" and all of the political effects that come from the scrum process

The only thing even worth discussing anymore is: what do you wish exits? Apps appear with no more than some words spoken into my phone. Dictation is the new way apps are built.

Self-improving native applications

Maintenance used to be an entirely separate part of the software development process. After an MVP, Beta, and version 1, the app would be handed of to a different set of people who would "maintain" the application, i.e. they would fix bugs that customers encountered while using the app. This was an entirely separate discipline that required specialized teams with unique talents. Customer support agents would triage issues that came in through customer service platforms. These platforms alone created businesses valued in the many billions of dollars. Hordes of employees would be requisitioned to support these applications and many would define their lives by what applications they supported. All of this existed to solve "bugs" in software. Entire schools of thought existed around software development methodologies on how to minimize the cost of these bugs, with assertions that production bugs were the most costly part of the development process.

Now, bugs can be fixed automatically, through simple error notifications. These notifications are transparent to the user. They generate solutions that appear almost before the process can be attempted again. Our software is becoming anti-fragile. This was a fable we aspired for. And it is here. We only need the awareness that it can be wished for and we can have it.

Software Engineering is dead

Many will argue, from a point of egocentricity, that software engineers are still needed specialists. They may point to the complexities of enterprise software. These SaaS applications were always a stop-gap. The web was built for text. It was made for simple documents, to share information between one party and another. It is remarkable that humans were able to make such a system work for applications. It was never supposed to be this way. Applications were meant to live on the computer. They were meant to exist inside of a platform.

The web platform is a breathing its last breaths. The web will go back to what it was always meant to be: a place where information is shared. It will cease to be a place where applications are distributed. It was easy to build a web app and point someone to an address. The messiness of the combination of different browsers and different operating systems, all running on different versions, meant that enormous amounts of energy were expended to make application backwards-compatible, cross-browser compatible. It never worked well. There were always browsers that caused issues. So we made frameworks. We made standards. We made an entire world of solutions, trying to resolve the problem that native apps were not easy to make, distribute, update. That time is over.

Applications will be built for a single platform, for the latest version of the operating system. If users wish to have a version work with their older system, they can use an older version of the app.

It is amazing how much work it has taken the software industry to make web applications work. I feel like it was all folly on some level. We had reached a point of local maxima and we could not see the world that could exist. But that world is here. Native is now easy, easier even than web. There are standard components that exist for native applications. The software industry created a myriad of frameworks, technologies, to build components that only worked with specific stacks. There were specialists for these stacks.

An engineer didn't build applications. React developers build React apps. An individual might specialize in a single library inside of the React ecosystem, like Redux. Highly respected individuals were awarded PHDs because they were able to write papers talking about how one can optimize the usage of these libraries. This is all evaporating in front of our eyes. The argument about Redux vs Mobx, vs whatever it's all pointless and moot now. Who cares! We've moved past it. Chuck it all in the bin. We've moved past an entire platform. The 10,000 things that made it up, the millions of individual careers worth of effort that has been put into this platform is obsolete.

The Bike Shed

There's a term called bike-shedding that has been used in software engineering for many years. It's the idea that groups can get stuck on trivial aspects of a project; that we can argue about the color of the bike-shed when we should be concerned with the actual factory. It seems like we've all been arguing about the bike shed for the last 20 years.